Download Sample 1

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Jane Pryma
Dr. B. Hartlaub
Elements of Statistics
24 April 2009
Statistical Analysis of Eating Disorders Within the Adolescent Ballet Community
The world of dance can be one of ruthless competition and unhealthy expectations. While
great strides have been made in changing some destructive standards of the business, there
continues to be a somber reality hiding behind the idealized beauty of, in particular, classical
dance forms. Ballet has long been the most evident perpetrator of negative body image among
dancers. Its taxing physical requirements paired with an aesthetic of long slender lines can easily
promote unhealthy body objectification among girls. By no means do all ballerinas entertain
injurious desires of unattainable body-shape, but there are, undeniably, many ballerinas that live
in constant fear of gaining weight, panicked that they will distort their body lines, ruin their
form, and thus, render themselves unable to perform professionally. This rationale appears
blatantly disparaging and unrealistic to an outsider, but to the dancer who lives to perform
through movement with a learned ethic of beauty, the fear of failure and inadequacy can easily
distort attitudes about eating.
In the 1998 article “The eating attitudes test: Comparative analysis of female and male
students at the Public Ballet School of Berlin” published in European Child and Adolescent
Psychology, the researchers conducted a study hypothesizing the notion, widely acknowledged in
the dance world, that female ballerinas or students of ballet are more apt to exhibit signs of an
eating disorder than their non-dancer peers.
Method
The researchers selected a stratified sample of adolescents age 11-18. Thirty-seven of
fifty-nine approached female ballet students agreed to take part in the study, and twenty of
twenty-two male ballet students agreed. The selected ballet-students approached were all
enrolled in the Public Ballet School of Berlin. The second group, which constituted the control
group, was comprised of Berlin high school students matched by age to the participants from the
Public Ballet School of Berlin. Ninety-five female students and sixty-one male students from the
same high school participated in the study. The third group consisted of hospitalized female
patients diagnosed with anorexia nervosa by the ICD-10 criteria for the disease. Nineteen
subjects were selected from this group. The mean age of female ballet students was 15.27 with a
standard deviation of 1.39. The mean age for male ballet students was 15.03 with a standard
deviation of 1.24. The mean age for the female high school student group was 15.52 with a
standard deviation of 1.43, and the mean age for their male counterparts was 14.96 with a
standard deviation of 1.26. The mean age for the female patients diagnosed with anorexia
nervosa was 14.89 with a standard deviation of 1.41. Thus, the three groups were relatively equal
in both mean and variance of age.
Each group was given the EAT-40 survey (Eating Attitudes Test), a standardized test that
measures the severity of an eating disorder or eating attitude. A test result exceeding 30 indicates
eating disorder. Each subject’s weight and height were measured in order to calculate each
participant’s BMI. The female participants were required to record their menstrual patterns. Of
specific interest to the study, was if menstruation was absent for female participants, and if so,
for how long. Mean values for the EAT-40 test results were interpreted using a t test, and the
distribution of EAT test results between two sample groups was evaluated with the MannWhitney U-test.
Results and Conclusion
The data collected were interpreted in four tables. Recording t values the researchers
arrived at these test statistic results from a two-sample t test for comparing two independent
groups’ means. The first table compares the mean EAT score for ballet students of both genders
to their respective control groups using t values. In the second table, the EAT data were broken
up into three subcategories. The EAT test is based on three factors. Through the participants’
survey responses the researchers were able to calculate the subscores for each factor, allowing
for further comparison between the eating attitudes of ballet dancers versus non-dancers. The
first factor is characterized by “avoidance of fattening food and preoccupation with being
thinner” (Neumärker et al. 22). “The second pertains to thinking about food and bulimic
behavior” (Neumärker et al. 22). The third factor relates to controlling one’s eating and a
“perceived pressure from others to gain weight” (Neumärker et al. 22). By comparing the results
within these subcategories, the researchers could compare the differences and similarities
between components that led to higher EAT scores. The third table records the means of total
EAT scores and the means of individual EAT factors for those females scoring thirty or above on
the EAT-40 test. This table includes result from all tested female groups: the ballet students, nondancing high school students and admitted patients. The fourth table lists the individual questions
from the EAT survey that represented significant differences between the girls in the tested
groups who scored a 30 or above on the EAT test.
With the data provided by the EAT surveys, the researchers first compared female
dancers with their complementary control group (female high school students) and the men
dancers with their complementary control group (male high school students.) The patients were
excluded in this comparison because the researchers were interested in determining if EAT
scores were higher for dancers compared to their presumably average, non-dancing peers. In this
situation the researches employed a two-sample t test. The null hypothesis for the female group
was that female dancers and non-dancers would have the same mean scores on the EAT test, and
the alternative hypothesis was that the female dancers would have a higher mean score on the
EAT test than their non-dancer peers. The data was collected and stratified by age group. With p
values lower than .05, the samples of females age sixteen to seventeen and the aggregate results
of females age thirteen suggest that the dancers statistically significant higher scores on the EAT
test. Thus, the researchers rejected the null hypothesis in favor of the hypothesis that the female
ballet students had a higher rate of anorexic behavior than their non-dancing peers. The t value
for the mean of the total sample of female ballet students minus the mean of the total sample of
female high school students was 2.78 with 130 degrees of freedom, and a p value of less than
0.005.
For males, similar data resulted. Mean EAT totals were higher for male ballet students
than their non-dancing counterparts. When interpreting the sum of all of the age groups, there is
strong statistical evidence that male ballet students have higher mean scores on the EAT test than
the male high school control group. The t value for the two sample t test comparing male
dancers’ EAT scores with the control group’s was 2.84 with 55 degrees of freedom and a p value
of less than 0.005
In Table 2, researchers determined that, for males, the overall higher mean EAT score
can be credited to a higher vale of Factor 3 that is statistically significant. When comparing the
mean scores of Factor 3 for male ballet dancers and male non-dancers, the two-sample t test
yielded a t value of 2.1, with 79 degrees of freedom and a p value of less than 0.05. Regarding
females in Table 2, female ballet students had statistically higher mean EAT scores for all three
factors compared to the control group of female non-dancers. Factor 1 resulted in a t score of
2.43 and a p valueless than 0.05. Factor 2 produced t value of 2.73 and a p value of 0.005, and
Factor 3 generated a t vale of 2.2 and a p value f less than 0.005. Thus, the differences were all
statistically significant, rejecting the null hypothesis for a two-sample t test that the non-dancers’
and dancers’ responses to the factors was the same. The dancers’ scores were significantly higher
than the control group. When comparing male and female ballet students with a two-sample t
test, the female dancers produced higher scores for Factors 1 and 2 with statistical significance
(for Factor 1, a t value of 3.87 with 55 degrees of freedom and a p value of less than .0005 was
achieved, and for Factor 2, a t value of 2.94 with 55 degrees of freedom and a p value of less
than 0.005 was achieved.) These differences between genders for the ballet student group were
also recorded for the high school student control group. Thus, the discrepancies between genders
appear to be consistent across groups.
In Table 3, the responses for the three factors between dancing females and non-dancing
females with EAT scores above 30 (indicating anorexia nervosa) appear similar. There is not
evidence to indicate a statistically strong difference between the two groups. The results of twosample t tests result in high p values, subsequently causing the researchers to not reject the null
hypothesis that mean values for the three factors are the same between female ballet students and
the female control group. However, in Factor 3, the female dancers and female control group
differed with statistical significance from the anorexia nervosa patients with a two-sample t test
between dancers and anorexia nervosa patients producing a t value of 6.55 with 25 degrees of
freedom, and for a two-sample t test between the control group and anorexia nervosa patients a t
value of 6.92 with 24 degrees of freedom and a p value of less than 0.0001 results.
With regards to Table 4, the article states, “Significant differences in response behaviour
were more frequently established between female ballet school students, on the one hand, and
female A.n. patients, on the other (eleven questions), than between the control group and female
patients (five questions)” (Neumärker et al. 21). However, the article does not provide test
statistics or p values to support this claim.
Based on the data, the researchers determined that, while mean EAT scores are higher for
ballet students than the control group, the ballet group does not statistically produce more
subjects with eating disorders than the control group. In fact, none of the ballet students or
students in the control group were anorexic based on ICD-10 criteria. Additionally, from the
responses to individual factors, the eating attitudes of female dancers were much closer to the
control group than to the female anorexic patients. The researchers concluded that the higher
mean EAT scores within the ballet student group were most likely a result of “their specific
living conditions” (Neumärker et al. 22). The researchers did not believe that the higher EAT
scores within the dancing population were not necessarily associated with psychological
disorders but rather a product of the culture. However, the researchers stressed that their study
did reaffirm the hypothesis that eating disorders are higher among females than males in
populations. Additionally, due to the data’s stratification by age, the researches asserted that
eating disorders seem to peak around age sixteen. Acknowledging the possibility of error, the
researchers contended, “the low specificity of the EAT when screening the general population,
the high evidence of elevated EAT scores within the control group compared with the reported
incidence of A.n…may support [the] conclusion that the female controls with suspicious EAT
scores rather were falsely positive cases” (Neumärker et al. 22).
Critique
I have some serious concerns with this study. First, the small sample size, in conjunction
with the selection of participants from two schools and one hospital does not bode well for
applying the results of this study to the general population. All of those tested lived in Berlin, the
ballet students all attended the same school, and the control group came from one high school.
The article itself states that the ballet school from which the participants were selected is one of
the premier ballet schools in Europe and is unique in its approach. This uniqueness is troubling. I
don’t think the results of these ballet students can be applied to ballet students everywhere. For
the results to be representative of the general population, the sample would need to be much
larger and include ballet students from many different schools from various parts of the world.
Similarly, the control group would need to be a more diverse population of high school age
students, and the anorexia nervosa patients would have to be selected from many different
hospitals from many different countries and regions.
There, too, is evidence of non-response bias. Not all of the approached ballet students
agreed to the study. Additionally, the reader is not informed of how the researchers selected the
participants comprising the control group or the anorexia nervosa group, leaving room to
question if the two samples were truly random.
It is also unnerving that the researchers admit to the possibility of attaining false positives
through the low specificity of the EAT survey when testing a large, average population. Perhaps
the researchers should not have used a test that has the possibility for producing false positives
when applied to general populations.
The researchers appear to have used accurate statistical methods. Their use of a twosample t test is appropriate, as is their utilization of p values in comparing inequalities between
groups. However, their last assertion that the data found in Table 4 shows strong differences in
response for individual factors between the female ballet students and anorexia nervosa patients
is unsupported by any reported p values.
Work Cited:
Neumärker, et al. "The eating attitudes test: Comparative analysis of female and male
students at the Public Ballet School of Berlin." European Child & Adolescent
Psychiatry 7.1 (1998), 19 - 23.